prompts.chatprompts.chat
Login
All classifiers
ScoreFeedback & research

Feedback sentiment

Rate how positive or negative a piece of feedback is, with neutral and mixed in the middle.

F@f

Problem

Star ratings don't match the text often enough to rely on, and a single positive/negative label hides mixed feedback.

Goal

Track sentiment per theme over time and find unhappy users worth a follow-up.

Input / state

The feedback text.

feedback
The feedback text.

Decision question

ScoreQuestion key sentiment

How positive or negative is this feedback overall?

Rubric levels

  • 0Very negative: angry or very disappointed.
  • 1Negative: unhappy or critical.
  • 2Neutral, or a mix of positive and negative.
  • 3Positive: satisfied or appreciative.
  • 4Very positive: enthusiastic.

Allowed outcomes

Decided in your code from the answer and its confidence. Thresholds are yours to set and test.

  • NegativeMost likely level is 0 or 1 and confidence meets your threshold.
  • Neutral or mixedMost likely level is 2 and confidence meets your threshold.
  • PositiveMost likely level is 3 or 4 and confidence meets your threshold.
  • UnscoredConfidence is below your threshold: probability is spread across options.

Example scenarios

Predefined demonstrations. No model is called; the outcome shown is an illustrative judgment.

Example state

{
  "feedback": "Lost a day of work because sync deleted my edits. Unacceptable."
}

Outcome

Illustrative
Negative

Most likely level is 0 or 1 and confidence meets your threshold.

Why

Angry and disappointed.

Using it well

  • Report averages per topic rather than overall; overall sentiment hides where the problems are.
  • The expected score across levels is a smoother trend measure than the most likely level.

Request definition

POST https://api.typesafe.ai/v1/systemone
{
  "model": "jev-latest",
  "state": {
    "feedback": "<feedback>"
  },
  "questions": {
    "sentiment": {
      "type": "score",
      "instructions": "How positive or negative is this feedback overall?",
      "criteria": [
        "Very negative: angry or very disappointed.",
        "Negative: unhappy or critical.",
        "Neutral, or a mix of positive and negative.",
        "Positive: satisfied or appreciative.",
        "Very positive: enthusiastic."
      ]
    }
  }
}

Request shape from the System One API reference; a Kev server accepts the same body. Replace the placeholder state with your data. Requests need your own API key or server.

Version History

v1Current Versionabout 8 hours ago@f

Initial version

Comments (0)